Date of Award

Fall 2014

Degree Type


Degree Name

Master of Science (MS)


Agricultural and Biological Engineering

First Advisor

Dennis Buckmaster

Second Advisor

Bernard Engel

Committee Member 1

Jane Frankenberger


With the advancement of mobile devices, opportunities to take watershed management tasks out of the office and into the field can be realized. In turn, field workers can utilize these technologies to expedite the decision-making process so that they may focus on meeting with clients and addressing agricultural watershed management issues. High-resolution (∼1.5 m postspacing) elevation data gathered by light detection and ranging (LiDAR) provides the topographic detail necessary to model hydrology at the field-scale (∼1 km2).

Non-artifactual surface depressions lead to erroneous surface flow patterns when using existing algorithms. So a sequential depression-filling algorithm (SDFA) has been developed to address topographies that contain these types of features. Given a rainfall amount, water distributed across the landscape accumulates and fills only those depressions as necessary, halting the filling process when the only depressions that remain require additional rainfall. After the filling process is completed, the watershed contributing area draining to any particular point of interest may be identified and in the future this may be used as input to hydrologic models. Methods have also been developed to implement subsurface drainage features such as culverts and tile-inlets as well as soil infiltration such that the dynamics of how water is shed from a given landscape can be better represented. Tile inlets and drainage features may be identified via user input and assigned a drainage rate while infiltration may be implemented by assigning a drainage rate to each grid cell in the DEM based on their soil-type. The combination of the sequential depression-filling algorithm and this drainage feature implementation provides the tools to model localized drainage patterns that will match user's field observations at the scale of hundreds of hectares.

The flow routing, depression identification, and filling procedures of the SDFA were compared to similar functions in the ArcGIS Hydrology Toolset under conditions where all depressions were filled in order to validate that those components of the algorithm are identical as intended. Furthermore, several digital elevation models (DEMs) were analyzed to determine the variability in hydrologic connectivity across these landscapes as a function of rainfall and as a function of DEM size. In addition to depression storage, the impacts of infiltration on hydrologic connectivity over these landscapes were also analyzed using the SCS Curve Number Method. The assumptions made by existing algorithms that require complete hydrologic connectivity do not hold up in all landscapes, even more so when considering the effects of infiltration. In these landscapes, surface hydrologic connectivity varies noticeably with rainfall excess and it is inaccurate to assume that the watershed should be modeled as a monotonically descending 14 surface. In an applicability study of DEM size, depression features began to be captured around the 1 km 2 scale while it is recommended to use DEMs larger than 2 km 2 to ensure that the depressional features and their contributing areas are completely captured within the DEM extent so that the SDFA may account for those features correctly.

The SDFA algorithm was ported from Matlab to an Android application for mobile phones and tablets. The Watershed Delineation app is free and publicly available through the Google Play Store. Users may view DEMs on a Google Map, use the sequential depression-filling algorithm to fill depressions, and delineate watersheds. It was found that the performance of this algorithm is a function of the number of depressions in the DEM which increases with DEM resolution (due to signal-noise effects). At a 3-meter resolution, the ideal DEM dimensions suitable for use of the SDFA on a Google Nexus 4 phone are about 500 x 500 (225 hectares), which took 68 seconds to run. At DEM sizes much greater than this, performance is drastically reduced. As DEM resolution increases, noise effects in the data (which vary based on the raw LiDAR data) result in a high amount of depression features causing an excessive number of iterations of the filling procedure within the algorithm.